The Agent Skills Directory

[DYNAMIC_EXECUTION]: The script scripts/render-range.py utilizes importlib.util to dynamically load the sibling script scripts/render-conversation.py at runtime. The file path is resolved statically relative to the main script's directory, which is a standard pattern for code sharing within a skill package.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from external conversation exports.
Ingestion points: External data enters the agent context through scripts/inspect-export.py, scripts/list-conversations.py, scripts/search-conversations.py, scripts/render-conversation.py, and scripts/render-range.py, which read contents from ZIP and JSON archives.
Boundary markers: The skill instructions do not specify any boundary markers or directives for the agent to ignore instructions embedded within the processed chat history.
Capability inventory: The scripts possess the capability to read local files (the archives) and write output to the file system via user-defined paths (e.g., /tmp/ or reference/chatgpt/). No network capabilities are present.
Sanitization: The skill extracts text using standard JSON and zip parsing; no specific sanitization or filtering is applied to identify or mitigate embedded instructions within the history content.

navigating-chatgpt-history